Multiple-Attention Mechanism Network for Semantic Segmentation

MFF-YOLO: An Accurate Model for Detecting Tunnel Defects Based on Multi-Scale Feature Fusion

Tunnel linings require routine inspection as they have a big impact on a tunnel's safety and longevity. In this study, the convolutional neural network was utilized to develop the MFF-YOLO model. To improve feature learning efficiency, a multi-scale feature fusion network was constructed within the neck network. Additionally, a reweighted screening method was devised at the prediction stage to address the problem of duplicate detection frames. Moreover, the loss function was adjusted to maximize the effectiveness of model training and improve its overall performance. The results show that the model has a recall and accuracy that are 7.1% and 6.0% greater than those of the YOLOv5 model, reaching 89.5% and 89.4%, respectively, as well as the ability to reliably identify targets that the previous model error detection and miss detection. The MFF-YOLO model improves tunnel lining detection performance generally.

Few-shot segmentation with duplex network and attention augmented module

Establishing the relationship between a limited number of samples and segmented objects in diverse scenarios is the primary challenge in few-shot segmentation. However, many previous works overlooked the crucial support-query set interaction and the deeper information that needs to be explored. This oversight can lead to model failure when confronted with complex scenarios, such as ambiguous boundaries. To solve this problem, a duplex network that utilizes the suppression and focus concept is proposed to effectively suppress the background and focus on the foreground. Our network includes dynamic convolution to enhance the support-query interaction and a prototype match structure to fully extract information from support and query. The proposed model is called dynamic prototype mixture convolutional networks (DPMC). To minimize the impact of redundant information, we have incorporated a hybrid attentional module called double-layer attention augmented convolutional module (DAAConv) into DPMC. This module enables the network to concentrate more on foreground information. Our experiments on PASCAL-5i and COCO-20i datasets suggested that DPMC and DAAConv outperform traditional prototype-based methods by up to 5-8% on average.

Fast Semantic Segmentation of Remote Sensing Images Using a Network That Integrates Global and Local Information

Efficient processing of ultra-high-resolution images is increasingly sought after with the continuous advancement of photography and sensor technology. However, the semantic segmentation of remote sensing images lacks a satisfactory solution to optimize GPU memory utilization and the feature extraction speed. To tackle this challenge, Chen et al. introduced GLNet, a network designed to strike a better balance between GPU memory usage and segmentation accuracy when processing high-resolution images. Building upon GLNet and PFNet, our proposed method, Fast-GLNet, further enhances the feature fusion and segmentation processes. It incorporates the double feature pyramid aggregation (DFPA) module and IFS module for local and global branches, respectively, resulting in superior feature maps and optimized segmentation speed. Extensive experimentation demonstrates that Fast-GLNet achieves faster semantic segmentation while maintaining segmentation quality. Additionally, it effectively optimizes GPU memory utilization. For example, compared to GLNet, Fast-GLNet's mIoU on the Deepglobe dataset increased from 71.6% to 72.1%, and GPU memory usage decreased from 1865 MB to 1639 MB. Notably, Fast-GLNet surpasses existing general-purpose methods, offering a superior trade-off between speed and accuracy in semantic segmentation.

CMANet: Cross-Modality Attention Network for Indoor-Scene Semantic Segmentation

Indoor-scene semantic segmentation is of great significance to indoor navigation, high-precision map creation, route planning, etc. However, incorporating RGB and HHA images for indoor-scene semantic segmentation is a promising yet challenging task, due to the diversity of textures and structures and the disparity of multi-modality in physical significance. In this paper, we propose a Cross-Modality Attention Network (CMANet) that facilitates the extraction of both RGB and HHA features and enhances the cross-modality feature integration. CMANet is constructed under the encoder-decoder architecture. The encoder consists of two parallel branches that successively extract the latent modality features from RGB and HHA images, respectively. Particularly, a novel self-attention mechanism-based Cross-Modality Refine Gate (CMRG) is presented, which bridges the two branches. More importantly, the CMRG achieves cross-modality feature fusion and produces certain refined aggregated features; it serves as the most crucial part of CMANet. The decoder is a multi-stage up-sampled backbone that is composed of different residual blocks at each up-sampling stage. Furthermore, bi-directional multi-step propagation and pyramid supervision are applied to assist the leaning process. To evaluate the effectiveness and efficiency of the proposed method, extensive experiments are conducted on NYUDv2 and SUN RGB-D datasets. Experimental results demonstrate that our method outperforms the existing ones for indoor semantic-segmentation tasks.